[cairo] [PATCH] Altivec optimizations

Baz brian.ewins at gmail.com
Thu Mar 8 19:39:23 PST 2007


On 09/03/07, Baz <brian.ewins at gmail.com> wrote:
> On 17/02/07, Luca Barbato <lu_zero at gentoo.org> wrote:
> > I completed the fbcompose part the regtest are consistent, hopefully
> > I'll start adding the pict ops later in the next week.
> >
> > lu
>
I'm seeing nearly zero difference in perf-diff (image surface). I
hacked in a check, altivec is definitely being enabled. I suspect
something wrong with perf-diff. The xlib surface is crashing, but its
unrelated to your patch, it crashes both before and after the patch on
that box. Most likely my fault, I'll take another run at it tomorrow.

When I say nearly zero difference, I mean literally zero unless I perf
-diff against several revisions ago (so that other things actually
make a difference), just to proved to myself that perf-diff is still
working. I've never seen it this consistent before, but there were
some changes in 1.4. Am I right in thinking the image backend should
be sufficient to demonstrate an effect of from your patch?

Cheers,
Baz


More information about the cairo mailing list